Classificador Ensemble: Uma Abordagem Não Paramétrica Aplicado à Detecção De Diabetes

When dealing with dichotomous data, that is, which admit only two possible answers, the commonly used supervised machine learning models are: logistic regression, random trees and K–Nearest neighbor (KNN). However, depending on the data, such predictions may not have a good accuracy (> 70%). As an alternative to the usual models, ensemble classifiers emerge (Gul & Perperoglou, 2018). The ensemble method is a machine learning technique that combines the results of multiple models in order to produce a better predictive model. There are several preset algorithms for ensemble classifiers, such as: bagging, boosting, bayesian averaging, among others. However, the choice of base predictive models and the way these results will be combined are free (Opitz & Maclin, 1999). The ensemble classifiers are a class of methods used to increase the accuracy of the model by joining weaker models, when the use of simpler methods, separately, does not present the desired result. For the application of more sophisticated methods for predicting the data, it is necessary to use initial techniques that allow the correct use of the model.